Hi all,
I built an open-source voice-command system for my PiDog V2 and wanted to share it here.
It runs entirely locally — speech-to-text with Whisper and intent classification with a small Ollama model (llama3.2) — so no cloud service is involved. Commands are defined in a JSON file, so you can add your own without touching the Python code. The robot barks once to acknowledge it heard a command, and twice when it didn’t understand.
Right now the wake word and commands are in French, but the JSON approach makes it straightforward to adapt to another language.
Happy to answer questions or take feedback.